/Tip of the iceberg – the US labour market catastrophe now playing out

Tip of the iceberg – the US labour market catastrophe now playing out

On April 3, 2020, the US Bureau of Labor Statistics (BLS) released their latest labour market data – Employment Situation Summary – March 2020 – which shows a deteriorating labour market situation due to the coronavirus crisis. However, as I explain below, the data released was drawn from samples that went up to March 12 (establishment survey) and March 14 (household survey), and so doesn’t fully capture the extent of the unfolding catastrophe. More recent data released by the US Department of Labor (unemployment insurance claimant data) doesn’t leave anything to doubt. In the last two weeks of March 2020, 9.955 million workers registered unemployment insurance claims (6.6 million in the last week). If we consider that shift, then the US unemployment rate would be around 9.8 per cent by the end of march and rising. All the aggregates are demonstrating dramatic shifts. The employment-population rate fell by 1.1 points to 60 per cent, which is the largest monthly fall since the sample began in January 1948. The U6 measure of broad labour underutilisation increased by 1.7 points to 8.7 per cent. This is the largest monthly rise in this measure since it was first published in January 1994. The situation will get worse.

Overview for March 2020:

  • Payroll employment fell by 701,000 – a sharp fall.
  • Total labour force survey employment fell by 2,987 thousand (net).
  • The seasonally adjusted labour force fell by 1,633 thousand (as participation fell).
  • Official unemployment rose by 1,353 thousand to 7,140 thousand (but see below for commentary on latest data)
  • The official unemployment rate rose from 3.5 to 4.4 per cent. But the actual March figure is close to 10 per cent (see below).
  • The participation rate plunged by 0.7 points to 62.7 per cent. Adjusting the unemployment rate for the change in hidden unemployment would generate an unemployment rate of 5.4 per cent rather than 4.4. per cent.
  • The broad labour underutilisation measure (U6) rose sharply by 1.7 points to 8.7 per cent because of a rise in the number in the part-time for economic reasons cohort (the US indicator of underemployment) and a rise in other marginal categories.

For those who are confused about the difference between the payroll (establishment) data and the household survey data you should read this blog post – US labour market is in a deplorable state – where I explain the differences in detail.

The catastrophic impacts of the health crisis – so far

The BLS provided a special explanatory note “Coronavirus (COVID-19) Impact on March 2020, Establishment and Household Survey Data” with their April 2020 labour market release.

We learned that the survey week – when the household survey was conducted ran from March 8 to March 14, 2020. This is when the unemployment estimates are derived.

The payroll employment data is derived from an establishment survey that spans “the pay period including the 12th of the month”.

Further, the conduct of both surveys was impacted by the virus. The labour force data (household survey) was exclusively conducted by telephone, which reduced the “response rate” to “73 percent … about 10 percentage points lower than in recent months”.

The BLS tell us in this – Technical Note – that:

At an unemployment rate of around 6.0 percent, the 90-percent confidence interval for the monthly change in unemployment as measured by the household survey is about +/- 300,000, and for the monthly change in the unemployment rate it is about +/- 0.2 percentage point.

So there is sampling error, which may have been larger than usual as a result of the reduced response rate.

However, the BLS also note that given the sample survey week ended on March 14, 2020:

We cannot precisely quantify the effects of the pandemic on the job market in March. However, it is clear that the decrease in employment and hours and the increase in unemployment can be ascribed to effects of the illness and efforts to contain the virus. It is important to keep in mind that the March survey reference periods for both surveys predated many coronavirus-related business and school closures in the second half of the month.

Already, up to March 14, 2020, there was an:

1. “extremely large increase in the number of persons classified as unemployed on temporary layoff.”

2. “there was also a large increase in the number of workers who were classified as employed but absent from work.”

The BLS instructed the interview team to count those employment but absent as “unemployed on temporary layoff”.

But they admit that:

… it is apparent that not all such workers were so classified. Such a misclassification is an example of nonsampling error and can occur when respondents misunderstand questions or interviewers record answers incorrectly.

What does that mean?

It means that the “the overall unemployment rate would have been almost 1 percentage point higher than reported.”

However, the BLS reports the data “as recorded”, which means they are above accusations of manipulating the data for political or other reasons.

So we know the unemployment situation is worse than reported in the official March data and this will be more clearly revealed in the April figures (released early May).

But we can speculate using the US Department of Labor’s Unemployment Insurance claims data, which is released on a weekly basis.

The Department of Labor provides an archive of the weekly unemployment insurance claims data back to July 1, 1967 – HERE.

The last two weeks data can be found in the – UI Weekly Claims Report.

The following reality emerges.

The Department of Labor reports that:

In the week ending March 28, the advance figure for seasonally adjusted initial claims was 6,648,000, an increase of 3,341,000 from the previous week’s revised level. This marks the highest level of seasonally adjusted initial claims in the history of the seasonally adjusted series.

Bringing together the archived data and the most recent release, the following table tells the shocking story.

Week ending Initial Claims (SA) Weekly Change 4-Week Moving Average (SA)
March 7, 2020 211,000 -6,000 161,500
March 14, 2020 282,000 +71,000 232,500
March 21, 2020 3,307,000 +3,025,000 1,004,250
March 28, 2020 6,648,000 +3,341,000 2,612,000

If we assume all those new claimants were unemployed then the unemployment level by the end of March would be 16,024 thousand instead of the BLS estimate (take up to the end of the survey week – March 14) of 7,140.

This would mean that the unemployment rate would be 9.8 per cent by the end of March rather than the official BLS March figure of 4.4 per cent.

The peak unemployment rate during the Great Depression was 24.9 per cent in 1933, before the New Deal brought it down somewhat.

The next two graph show the extraordinary nature of the current period.

The first graph shows new weekly claimants for unemployment benefits from July 1, 1967 to the week-ending March 13, 2020 (that is two weeks ago).

You can see the big cycles in the US labour market – the mid-1970s downturn, the recessions in the early 1980s, the early 1990s recession, the early 2000s and then the GFC.

The next graph adds the last two weeks of available data (week-ending March 21 and March 28) to the entire sample.

The spike at the end of the graph shows how drastic the situation is in the US.

As a researcher who has dealt with time-series data all my career, I have never seen anything quite that stark in a dataset.

Payroll employment trends

The BLS noted that:

Total nonfarm payroll employment fell sharply in March (-701,000), reflecting the effects of the coronavirus and efforts to contain it. About two-thirds of the drop occurred in leisure and hospitality, mainly in food services and drinking places. Notable employment declines also occurred in health care and social assistance, professional and business services, retail trade, and construction. In the prior 12 months, nonfarm employment growth had averaged 196,000 per month.

The first graph shows the monthly change in payroll employment (in thousands, expressed as a 3-month moving average to take out the monthly noise). The gray lines are the annual averages.

This month saw the rapid fall in payroll employment – which really blows out any comparison with past performance.

For the time being, this sort of plunge will be the norm.

The next graph shows the same data in a different way – in this case the graph shows the average net monthly change in payroll employment (actual) for the calendar years from 2005 to 2020 (the 2020 average obviously is just the January result).

The red diamond is the current month’s observation.

The slowdown that began in 2015 continued through 2017 was reversed last year. The 2018 average was 223 thousand compared to 179 thousand in 2017.

The final average for 2019 was 178 thousand.

While it looked like 2020 was started well in terms of quantity of jobs, a rise in the average rise for the first two months, the March result has changed things dramatically.

To put the current recovery into historical perspective the following graph shows the average annual growth in payroll employment since 1960 (blue columns) with the decade averages shown by the red line.

It reinforces the view that while payroll employment growth has been steady since the GFC ended, it was still well down on previous decades of growth before the latest crisis hit.

Labour Force Survey – employment growth collapses

The BLS report that:

In March, the unemployment rate increased by 0.9 percentage point to 4.4 percent. This is the largest over-the-month increase in the rate since January 1975, when the increase was also 0.9 percentage point. The number of unemployed persons rose by 1.4 million to 7.1 million in March. The sharp increases in these measures reflect the effects of the coronavirus and efforts to contain it.

The data for March is staggering:

Employment as measured by the household survey fell by 2,987 thousand net (-1.88 per cent) while the labour force declined by 1,633 thousand (-0.99 per cent) as the participation rate fell by 0.7 points (a huge monthly contraction.

As a result (in accounting terms), total unemployment rose by 1,353 thousand and the unemployment rate rise by 0.9 points to 4.4 per cent.

However, as the previous analysis showed this is just the tip of the iceberg.

For the time being the extended period of low unemployment is over.

The next graph shows the monthly employment growth since January 2008. The red line is the average labour force growth over the period December 2001 to December 2006 (0.09 per cent per month).

The summary conclusion: As in the previous graph (UI claimants), the decline in employment growth is dramatic – much worse than the GFC.

The Employment-Population ratio shift is also dramatic.

This is a good measure of the strength of the labour market because the movements are relatively unambiguous because the denominator population is not particularly sensitive to the cycle (unlike the labour force).

The following graph shows the US Employment-Population from January 1948 to March 2020.

While the ratio fluctuates a little, the March 2020 ratio fell by 1.1 points to 60 per cent, which is the largest monthly fall since the sample began in January 1948.

Unemployment and underutilisation trends

The first graph shows the official unemployment rate since January 1950 which shifted from 3.5 per cent to 4.4 per cent (noting the caveats at the outset.

The official unemployment rate is a narrow measure of labour wastage, which means that a strict comparison with the 1960s, for example, in terms of how tight the labour market, has to take into account broader measures of labour underutilisation.

The next graph shows the BLS measure U6, which is defined as:

Total unemployed, plus all marginally attached workers plus total employed part time for economic reasons, as a percent of all civilian labor force plus all marginally attached workers.

It is thus the broadest quantitative measure of labour underutilisation that the BLS publish.

In December 2006, before the effects of the slowdown started to impact upon the labour market, the measure was estimated to be 7.9 per cent.

In March 2020 the U6 measure increased by 1.7 points to 8.7 per cent.

This is the largest monthly rise in this measure since it was first published in January 1994.

This was driven, in part, by the rise in the category ‘Part-time for economic reasons’ – a measure of underemployment in the US data context) – it rose by a huge 1,447 thousand or 25.1 per cent.

The BLS say that “These individuals … would have preferred full-time employment, were working part time because their hours had been reduced or they were unable to find full-time jobs”.

So not only are there massive layoffs occurring but firms are also making hours adjustment – cutting full-time hours back to part-time or cutting part-time hours.

There was also a sharp increase in the category – Discouraged NLF search and available not available – which taken together with the rise in underemployment compounds the loss of jobs and rise in official unemployment.

The U-6 measure is now below the pre-GFC level, and, while, signalling improvement, there is still some scope to go before full capacity is reached.

Aggregate participation rate – plunges by 0.by 0.1 points to 66 per cent

The fall in the labour force participation rate eased the pressure on official unemployment.

By how much would unemployment have risen if the participation rate had not fallen?

The labour force is a subset of the working-age population (those above 15 years old). The proportion of the working-age population that constitutes the labour force is called the labour force participation rate. Thus changes in the labour force can impact on the official unemployment rate, and, as a result, movements in the latter need to be interpreted carefully. A rising unemployment rate may not indicate a recessing economy.

The labour force can expand as a result of general population growth and/or increases in the labour force participation rates.

The labour force would have been 164,732 thousand instead of the actual official value of 162,914 thousand had the participation rate not fallen.

If we add the difference back into the official unemployed then we would get an adjusted unemployment rate of 5.4 per cent instead of 4.4 per cent.


The March 2020 US BLS labour market data release reveals a rapidly deteriorating labour market.

But as the song goes – You ain’t seen nothing yet!

The weekly claimant data released in the last two weeks of March (after the survey end-date for the household survey) shows how dramatic the deterioration revealed in the official BLS data is much worse.

Once these trends are factored into the next month’s labour force survey then the unemployment rate will jump up past 10 per cent.

All variables reported changed sharply.

I have never seen these sorts of monthly shifts.

The scale of the crisis requires dramatic interventions.

The current position of the US government has biased their stimulus spending towards the corporate sector and high income earners.

It disadvantages precarious workers and those losing jobs.

The US government should immediately announce an unconditional Job Guarantee and guarantee salaries for all workers made redundant.

That is enough for today!

(c) Copyright 2020 William Mitchell. All Rights Reserved